Anybody tried the BerkeleyDB 3.2.3c version with -CURRENT?

[Heath Kehoe]

>
>On 23 Nov 2000, Jaye Mathisen wrote:
>
>> Was snagging the 3.2.3c release for mysql, and was contemplating giving it
>> a go on INN, on the assumption that more bugs fixed, and all that stuff.
>> Just curious if anybody else has tried it.
>
>Okay, ``Yes''.
>
>One needs to make about a half-dozen line hack to add ->set_flags for
>the NOSYNC transaction flag, rather than passing it as a flag to ->open,
>and it doesn't seem to like the checkpoint logs left behind by the old
>BerkeleyDB, but after I blew them out of the way, it's been running
>smoothly.

I just pulled down db-3.2.3c from the mysql site.  The docs indicate
that the above change is all that's necessary to make it work.
I will get the necessary modification into -current soon.


>
>It even made it through the point where 3.1.17 had been regularly
>falling over, expireover, but unfortunate netwerk problems caused the
>incoming feed and readers to disappear during this time, so I'm not
>sure if it's a reasonable test.  Further, I've done initial tuning
>for ovdb by offloading the checkpoint logs onto a separate disk, if
>the beating that the overview data gets during expireover has any
>relation to the mysterious failure of the logremoval procedure.
>

If logs still aren't getting removed, that would warrant an
examination of ovdb_monitor, to make sure it's really doing what
it's supposed to do.  Note that the checkpoint portion of
ovdb_monitor could be at fault.  A log file can't be removed until
all of the transaction log entries in it have been checkpointed.

If you notice that logs aren't being removed, try running:
   /berkdb/path/bin/db_checkpoint -h /path/overview -1


>
>And the other machine that isn't even *doing* anything keeps falling
>over by itself...  Furrfu...
>
>
>Disclaimer:  I have no idea what I'm doing, but I'd be happy to post
>the patches needed to be able to use 3.2.3c if you want to try it, if
>you are in fact still seeing problems with 3.1.17 or whatever.  Also,
>less than 24 hours of non-stressful testing is not enough to say that
>it is free of problems, so if I start to have them, I'll be eating my
>words.  Smothered in wasabi, mmmm
>

The MySQL folks must have a support arrangement with Sleepycat to
get newer BerkeleyDB codez than the public releases.  Which makes
sense, as the MySQL folks are probably fixing BerkeleyDB bugs.

I'll do some testing of 3.2.3 as well, in the next few days.

-h